| 摘要 |
A comparison function based OCR method is employed for recognizing the nearest neighbor classification of an input vector S with respect to a library of T template vectors. The input vector S defines an unclassified symbol (the unknown input letter) and the T template vectors define the pre-classified characters (known letters of an alphabet font in memory). Nearest neighbor classification is an iterative OCR matching technique in which the closest match between the input vector and the template vectors is recognized by the exclusion of distant templates. The closest match (the nearest neighbor) is accepted as the proper identification of the input symbol. The OCR method is accelerated by employing the Triangle of Inequality Principle which excludes distant neighbors in bulk as the iteration cycles proceed. The sequence of selection of template vectors from the diminishing candidate group may be random or ordered. Ordered selections optimize the exclusion process resulting in a nearest neighbor classification of the input symbol with fewer iterations. The ordering may be based on the frequency of use of the template characters, or on "historical inference", the statistical likelihood that any known character will be followed (or preceded) by any particular character remaining in the candidate group. <IMAGE> |
